Implementing Offset-Based Pagination

Make sure to check our introduction to pagination here !

Let us implement offset-based pagination in GraphQL. By the end of this tutorial, we should be able to return a sorted, filtered, and paginated list of users.

Let us model the User type, which represents one user, with a name, occupation, and age.

from typing import List, TypeVar, Dict, Any, Generic
import strawberry


@strawberry.type
class User:
    name: str = strawberry.field(description="The name of the user.")
    occupation: str = strawberry.field(description="The occupation of the user.")
    age: int = strawberry.field(description="The age of the user.")

    @staticmethod
    def from_row(row: Dict[str, Any]):
        return User(name=row["name"], occupation=row["occupation"], age=row["age"])

Let us now model the PaginationWindow , which represents one “slice” of sorted, filtered, and paginated items.

Item = TypeVar("Item")


@strawberry.type
class PaginationWindow(Generic[Item]):
    items: List[Item] = strawberry.field(
        description="The list of items in this pagination window."
    )

    total_items_count: int = strawberry.field(
        description="Total number of items in the filtered dataset."
    )

Note that PaginationWindow is generic - it can represent a slice of users, or a slice of any other type of items that we might want to paginate.

PaginationWindow also contains total_items_count , which specifies how many items there are in total in the filtered dataset, so that the client knows what the highest offset value can be.

Let’s define the query:

@strawberry.type
class Query:
    @strawberry.field(description="Get a list of users.")
    def users(
        self,
        order_by: str,
        limit: int,
        offset: int = 0,
        name: str | None = None,
        occupation: str | None = None,
    ) -> PaginationWindow[User]:
        filters = {}

        if name:
            filters["name"] = name

        if occupation:
            filters["occupation"] = occupation

        return get_pagination_window(
            dataset=user_data,
            ItemType=User,
            order_by=order_by,
            limit=limit,
            offset=offset,
            filters=filters,
        )


schema = strawberry.Schema(query=Query)

Now we’ll define a mock dataset and implement the get_pagination_window function, which is used by the users query.

For the sake of simplicity, our dataset will be an in-memory list containing four users:

user_data = [
    {
        "id": 1,
        "name": "Norman Osborn",
        "occupation": "Founder, Oscorp Industries",
        "age": 42,
    },
    {
        "id": 2,
        "name": "Peter Parker",
        "occupation": "Freelance Photographer, The Daily Bugle",
        "age": 20,
    },
    {
        "id": 3,
        "name": "Harold Osborn",
        "occupation": "President, Oscorp Industries",
        "age": 19,
    },
    {
        "id": 4,
        "name": "Eddie Brock",
        "occupation": "Journalist, The Eddie Brock Report",
        "age": 20,
    },
]

Here’s the implementation of the get_pagination_window function. Note that it is generic and should work for all item types, not only for the User type.

def get_pagination_window(
    dataset: List[GenericType],
    ItemType: type,
    order_by: str,
    limit: int,
    offset: int = 0,
    filters: dict[str, str] = {},
) -> PaginationWindow:
    """
    Get one pagination window on the given dataset for the given limit
    and offset, ordered by the given attribute and filtered using the
    given filters
    """

    if limit <= 0 or limit > 100:
        raise Exception(f"limit ({limit}) must be between 0-100")

    if filters:
        dataset = list(filter(lambda x: matches(x, filters), dataset))

    dataset.sort(key=lambda x: x[order_by])

    if offset != 0 and not 0 <= offset < len(dataset):
        raise Exception(f"offset ({offset}) is out of range " f"(0-{len(dataset) - 1})")

    total_items_count = len(dataset)

    items = dataset[offset : offset + limit]

    items = [ItemType.from_row(x) for x in items]

    return PaginationWindow(items=items, total_items_count=total_items_count)


def matches(item, filters):
    """
    Test whether the item matches the given filters.
    This demo only supports filtering by string fields.
    """

    for attr_name, val in filters.items():
        if val not in item[attr_name]:
            return False
    return True

The above code first filters the dataset according to the given filters, then sorts the dataset according to the given order_by field.

It then calculates total_items_count (this must be done after filtering), and then slices the relevant items according to offset and limit .

Finally, it converts the items to the given strawberry type, and returns a PaginationWindow containing these items, as well as the total_items_count .

In a real project, you would probably replace this with code that fetches from a database using offset and limit .

Tip

If you’re using Strawberry with the Django web framework, you might want to make use of the Django pagination API. You can check it out here .

Running the Query

Now, let us start the server and see offset-based pagination in action!

strawberry server example:schema

You will get the following message:

Running strawberry on http://0.0.0.0:8000/graphql 🍓

Go to http://0.0.0.0:8000/graphql to open GraphiQL, and run the following query to get first two users, ordered by name:

{
  users(orderBy: "name", offset: 0, limit: 2) {
    items {
      name
      age
      occupation
    }
    totalItemsCount
  }
}

The result should look like this:

{
  "data": {
    "users": {
      "items": [
        {
          "name": "Eddie Brock",
          "age": 20,
          "occupation": "Journalist, The Eddie Brock Report"
        },
        {
          "name": "Harold Osborn",
          "age": 19,
          "occupation": "President, Oscorp Industries"
        }
      ],
      "totalItemsCount": 4
    }
  }
}

The result contains:

items - A list of the users in this pagination window
totalItemsCount - The total number of items in the filtered dataset. In this case, since no filter was given in the request, totalItemsCount is 4, which is equal to the total number of users in the in-memory dataset.

Get the next page of users by running the same query, after incrementing offset by limit .

Repeat until offset reaches totalItemsCount .

Running a Filtered Query

Let’s run the query again, but this time we’ll filter out some users based on their occupation.

{
  users(orderBy: "name", offset: 0, limit: 2, occupation: "ie") {
    items {
      name
      age
      occupation
    }
    totalItemsCount
  }
}

By supplying occupation: "ie" in the query, we are requesting only users whose occupation contains the substring “ie”.

This is the result:

{
  "data": {
    "users": {
      "items": [
        {
          "name": "Eddie Brock",
          "age": 20,
          "occupation": "Journalist, The Eddie Brock Report"
        },
        {
          "name": "Harold Osborn",
          "age": 19,
          "occupation": "President, Oscorp Industries"
        }
      ],
      "totalItemsCount": 3
    }
  }
}

Note that totalItemsCount is now 3 and not 4, because only 3 users in total match the filter.